We recruited 24 male students (mean age, 22 years; range, 21–25 years; standard deviation, 1.39), and five were excluded from the statistical analysis because of technical problems and data loss. Therefore, 19 participants were included in the data analysis. The inclusion criteria were as follows: (1) healthy young male with normal uncorrected or corrected vision; (2) right-handed; (3) no use of medications or drugs, including psychotropic drugs, sleeping pills, cold medicines, etc., in the 2 weeks before the experiment; (4) no serious medical history such as hepatitis, cancer, nephritis, diabetes, or endocrine disorders; (5) no history of psychiatric diseases; (6) no habit of drinking coffee, tea, or smoking; (7) had not taken caffeinated or alcoholic beverages and had not exercised vigorously for 2 weeks prior to the experiment; (8) Pittsburgh Sleep Quality Index Questionnaire (Buysse et al., 1989) scores before the experiment <5; and (9) did not participate in psychological and physiological tests before. All participants signed written informed consent prior to the experiment, and our study was approved by the Biological and Medical Ethics Committee of Beihang University (ethics code: BM20180040).

Participants were asked to report their bedtime and waking times each day for 2 weeks prior to the experiment to ensure a regular sleep duration of 7–9 h. They arrived at the laboratory at 18:00 on the first day of the study period (Figure 1). Then, they were informed of the procedure and risks of the experiment and then signed a written informed consent form. The participants were asked to sleep in our laboratory for the night prior to the formal experiment. The formal experiment began at 8:00 the next day. Each participant underwent 36 h of TSD, and all the data were acquired in four experimental conditions [two eye states (EC; EO) and two sleep deprivation times (RW: 0 h; TSD: 36 h)]. Participants were asked to sit quietly and relax for 2 min upon wearing the EEG equipment. All the participants underwent the resting-state EEG tasks both in EC and EO states, and the different eye states were counterbalanced. Twelve participants were asked to keep their eyes open for the first 5 min, while the other twelve kept their eyes closed for the first 5 min, with a 2-min break in between.

They completed the first resting-state EEG task and PVT successively, meanwhile, behavioral and EEG data were recorded. After TSD (at 20:00 on the third day), the second resting-state EEG task and PVT were completed. During the whole experiment time, participants were allowed to have meals only at 7:30, 12:00, and 17:30. In the rest of the experimental period, participants could conduct their personal affairs. Simultaneously, they were supervised by our main staff to prevent them from napping. Meanwhile, vigorous exercise and high-arousal activities were forbidden, and they could not leave the laboratory. When completing the experiment, participants had recovery sleep in our laboratory and left the laboratory subsequently.

The PVT is highly sensitive to SD and has become the most widely used measurement approach for behavioral vigilance (Dorrian et al., 2005; Lim and Dinges, 2008). Thus, it was selected to indicate the participants’ psychomotor vigilance levels. At the beginning of each trial, the symbol “+” appeared for 400 ms at the center of a white background on the display screen (1,024 × 768-pixel resolution; refresh rate, 60 Hz). Then, red dots (diameter: 3 cm; viewing angle: 1.5 × 1.5°) appeared at the center of the screen for 1,000 ms and immediately disappeared when the participants pressed the button. The intervals between the trials were 8–12 s (mean: 10 s) and were pseudorandomized. Participants were asked to react as soon as a red dot appeared; however, early responses were not allowed. There were 30 trials in total; reaction times (RTs) to trials ≤150 ms were defined as an early response, and RTs ≥550 ms were defined as a lapse.

The EEG data were recorded based on the international standard 10–20-electrode-placement system using Neuroscan 32-channel electrode caps combined with a Syn-Amps2 amplifier (Compumedics Neuroscan, Charlotte, NC, USA) at a 1000-Hz sampling rate. All electrode impedances were < 5 kΩ prior to recording. Two electrodes were placed outside both eyes to record the horizontal electrooculogram, and another two electrodes were placed on the upper and lower sides of the left eye to record the vertical electrooculogram. The data of five participants were excluded from the statistical analysis due to data loss. Hence, the final sample comprised 19 participants.

The EEGLAB toolbox in MATLAB R2020B software (Mathworks, Natick, MA, USA) was used to preprocess the raw EEG signals offline.

In this study, we employed eLORETA to generate the estimated cortical electrical distribution from scalp signals because scalp signals cannot be directly used to compute cortical connections (Pascual-Marqui and Biscay-Lirio, 2011). The EEG signal is formed by the firing of neurons, and it reaches the scalp electrode through the tissue media such as cerebrospinal fluid and skull. Scalp potentials are affected by volume conduction effects so that electrodes at a given scalp location can detect the activity of neurons in the area directly below them as well as the activity of a remote source. The activation of one source simultaneously affects all scalp electrodes, resulting in an intrinsic correlation between the signals recorded on these electrodes. Therefore, signals acquired on scalps cannot be directly used to compute cortical connections, and the analysis of recorded EEG signals at different locations to determine the source of their generation is referred to as source localization analysis.

We opted to use eLORETA because it is an improved version of both LORETA and standardized LORETA (Pascual-Marqui, 2002). According to Pascual-Marqui, eLORETA is a weighted minimum norm inverse solution that provides exact localization with zero error, even in the presence of measurement and structured biological noise (Pascual-Marqui, 2007). The solution space comprised 6239 cortical gray matter voxels at a spatial resolution of 5 mm in a realistic head model (Fuchs et al., 2002). The Montreal Neurologic Institute (MNI) 152 template was used with Brodmann area (BAs) anatomic labels (Mazziotta et al., 2001).

Through functional independent component analysis, we aimed to examine the changes of spontaneous brain activity. Resting-state EEG data were analyzed by functional independent component analysis (fICA) using eLORETA to compute changes in neural oscillations (Pascual-Marqui and Biscay-Lirio, 2011). Cortical electrical activity was calculated for the following frequency bands: 1 Hz < δ ≤ 4 Hz; 4 Hz < θ ≤ 8 Hz; 8 Hz < α ≤ 12 Hz; 12 Hz < β ≤ 30 Hz; 30 Hz < γ ≤ 40 Hz. EEG data were converted into frequency domain to generate a set of cross spectral indicators using discrete Fourier transform. Subsequently, the spectral density of each frequency band and cortical voxel was calculated. Networks were divided according to the similarity of signals, in which similar signals were assigned to the same network, and the dissimilar signals were divided into different networks. Therefore, we obtained five networks. There were five coefficients corresponding to the five networks, with each representing a different frequency band (δ, θ, α, β, and γ). It is important to note that the cortical electrical distribution of this network was quite different from that of known functional networks (Chen et al., 2013). Functional networks deriving from fICA were data-driven, and the figures exported by eLORETA were not presented as a known whole functional network, and the brain regions of DMN and SMN might be mixed. Next, each participant’s data were linked to produce a matrix with dimensions of spatial frequency and different participants (Aoki et al., 2015).

Subsequently, the five coefficients were statistically analyzed to compare the differences in the functional networks of the resting states between different eye states or different SD times. Paired t-tests w ere applied to test RW (EO) = RW (EC), TSD (EO) = TSD (EC), TSD (EO) = RW (EO), TSD (EC) = RW (EC), and TSD (EO-EC) = RW (EO-EC). Simultaneously, randomization statistical nonparametric mapping (SnPM) was used to correct multiple comparisons (Thatcher et al., 2005; Painold et al., 2020). Finally, the functional networks with significant differences were identified, which were represented by independent components (ICs). They were permitted to have minor differences and overlap with traditional functional networks. They were not presented as a typical DMN or SMN, and the changed activity in each frequency band was presented as a network.

Functional connectivity has been extensively used to analyze functional networks. As a special functional connectivity, the effective connectivity provides direction information regarding the connections between different brain regions. We employed the effective connectivity analysis to assess the changes in synchronous activity in different conditions. Effective connectivity analysis was performed using eLORETA software with the isolated effective coherence (iCoh) algorithm. The definition of iCoh is based on formulating a multivariate autoregressive model and calculating the corresponding partial coherences after setting all irrelevant connections to zero other than the particular directional association of interest (Pascual-Marqui et al., 2014). We aimed to assess the direct paths of intracortical causal information flow of oscillatory activity within the SMN and DMN and the related changes between different eye states or different SD times. First, we selected brain regions in the SMN and DMN as the regions of interest (ROIs) by referring to previous studies and the results of fICA (Table 1). Then, iCoh values were calculated for 1–40 Hz, corresponding to five frequency bands: 1 Hz < δ ≤ 4 Hz; 4 Hz < θ ≤ 8 Hz; 8 Hz < α ≤ 12 Hz; 12 Hz < β ≤ 30 Hz; 30 Hz < γ ≤ 40 Hz. Paired t-tests were used to test RW (EO) = RW (EC), TSD (EO) = TSD (EC), TSD (EO) = RW (EO), TSD (EC) = RW (EC), and TSD (EO-EC) = RW (EO-EC). Randomization SnPM was used (number of randomizations: 5000) for the multiple comparison correction (Painold et al., 2020), then we acquired the corrected critical thresholds and p-values.

We used SPSS software (version 26.0) to perform paired t-tests on the PVT parameters between RW and TSD. Pearson’s correlation analysis was used on the changes in significant effective connectivity and PVT parameters before and after TSD. The level of statistical significance was set at p < 0.05. We corrected multiple comparisons using the False Discovery Rate (FDR). We chose Pearson’s correlation because it was the most common and effective way to measure a linear relationship between two continuous variables. Examining the correlations between effective connectivity and PVT response time could better illustrate the physiological significance of effective connectivity.

We chose paired t-tests as our statistical method for the fICA and iCoh analyses because we were not concerned with the interaction effect between our two independent variables; we were only concerned with the comparison between two states (EC and EO; RW and TSD). Moreover, after the t-test, we had strictly corrected the value of p to rule out false positive results by using SnPM. As our statistical method was able to verify the research hypotheses, we did not use analysis of variance.

The PVT parameters of the 19 participants differed significantly between the RW and TSD groups (Table 2). The paired t-test results suggested that compared with RW, the mean RT was significantly longer after TSD (t₁₈ = 2.31, P_{FDR-corrected} = 0.041). Furthermore, the median RT was significantly longer (t₁₈ = 3.07, P_{FDR-corrected} = 0.023), the fastest 10% RT was significantly slower (t₁₈ = 2.15, P_{FDR-corrected} = 0.045), the lowest 10% RT was significantly prolonged (t₁₈ = 2.37, P_{FDR-corrected} = 0.041), and the number of lapses was significantly higher (t₁₈ = 2.91, P_{FDR-corrected} = 0.023) after TSD. We corrected multiple comparisons using False Discovery Rate (FDR).

The fICA results showed that the independent components (ICs), representing data-driven functional network, significantly differed in the EC and EO states. The neural oscillations in the SMN were significantly different in the EO state compared to the EC state. This was manifested in the postcentral gyrus and superior parietal lobule, with reduced neural oscillations in the δ and θ bands and enhanced oscillations in the α and β bands in the EO state (Figures 2–4). Additionally, the neural oscillations in the DMN were significantly decreased in the EO state compared to the EC state. This was reflected in the PCu in the δ and θ bands (Figures 2, 3).

The fICA results showed that in the EC state, two ICs significantly differed between RW and TSD. In the EC state, neural oscillations in the PCu significantly decreased in the δ band in TSD compared with RW (Figure 5). Moreover, neural oscillations of the β band (Figure 5) in the SMN significantly increased, including in the postcentral gyrus (BA7, 5, 2, 1, 3, and 40), superior parietal lobule (BA7), and precentral gyrus (BA4 and 6).

The iCoh results suggested that the effective connectivity in the DMN in the resting-state was significantly higher in the EO state than in the EC state (t-corrected = 4.046, p-corrected = 0.0016) (Figure 6). The right IPL sent to the left MFG, oscillating at the α and β bands (t-corrected = 4.453); the right PCC sent to the right PCu oscillating at the α and β bands (t-corrected = 4.926), and sent to the right IPL oscillating at the β and γ bands (t-corrected = 5.510).

In addition, the iCoh results indicated that the effective connectivity in the SMN was significantly enhanced in the resting-state with EO compared to that with EC (t-corrected = 3.900) (Figure 6). The right SPL sent to the right SFG oscillating at α, β, γ bands (t-corrected = 6.097), and to the left SFG at β and γ bands (t-corrected = 5.149); the right PreG sent to the right (t-corrected = 4.868) and left (t-corrected = 4.733) SFG oscillating at α, β and γ bands, and to the left PreG at β and γ bands (t-corrected = 4.264); the left SFG sent to the right SFG oscillating at the γ band (t-corrected = 4.623); the right SFG sent to the left SPL oscillating at α and β bands (t-corrected = 4.449), to the left SFG at β band (t-corrected = 4.331), to the left PreG at β and γ bands (t-corrected = 4.292), and to the left PostG at α and β bands (t-corrected = 4.286); the right PostG sent to the right SFG oscillating at γ band (t-corrected = 4.714). Among all the frequency bands, the β band was the most significant for above effective connectivity in DMN and SMN.

The iCoh results suggested that the effective connectivity in the DMN was significantly enhanced in EO compared with EC in TSD (t-corrected = 4.034, p-corrected = 0.0092) (Figure 7), including the effective connectivity from the left MFG to the left (t-corrected = 4.675) and right PCC (t-corrected = 4.277) in δ, θ, and α bands; from the right MFG to the left PCC (t-corrected = 4.501) in α and β bands and to the right PCC (t-corrected = 4.214) in α and β bands; from the right PCC to the left PCu (t-corrected = 4.188) in δ band, to the right PCu (t-corrected = 4.183) in δ and θ bands, and to the left PCC (t-corrected = 4.568) in δ, θ and α bands.

The iCoh results suggested that the effective connectivity in the SMN was significantly enhanced in EO compared with EC in TSD (t-corrected = 3.878, p-corrected < 0.001) (Figure 7), including the effective connectivity from the right SPL to right SFG in θ, α, β, and γ bands (t-corrected = 6.842), to the left SFG in α, β and γ bands (t-corrected = 6.813), to the right PreG in δ, θ, α, β, and γ bands (t-corrected = 5.783), to the left PreG in γ band (t-corrected = 4.159); from the right PostG to the right (t-corrected = 5.095) and left (t-corrected = 4.711) SFG in β, and γ bands, to the left PreG in γ band (t-corrected = 3.902); from the left SPL to the left SFG in β band (t-corrected = 3.909). Among all the frequency bands, the γ band was the most significant for above effective connectivity in SMN.

The iCoh results suggested that in the EC state, the effective connectivity from the left PCu to the right PCu was significantly reduced in the β band after TSD (t-corrected = 4.041, p-corrected = 0.048) (Figure 8). Conversely, the effective connectivity from the right precentral gyrus to the left superior frontal gyrus was significantly enhanced in the β band after TSD in the EO state (t-corrected = 3.942, p-corrected = 0.036) (Figure 8).

The iCoh results suggested that the effective connectivity in the different networks (EO-EC) of the SMN was significantly decreased after TSD compared to RW (t-corrected = 4.046, p-corrected = 0.0016) (Figure 9). The left SPL sent to the left PreG oscillating at the β and γ bands (t-corrected = 3.5276, p-corrected < 0.049) and the left PostG sent to the left PreG oscillating at the γ band (t-corrected = 4.065, p-corrected < 0.049).

Pearson’s correlation analysis showed that after TSD, there was a significant positive correlation between the increase in the fastest 10% RT during the PVT and the decrease in the effective connectivity from the left PCu to the right PCu (r = 0.477, p = 0.039) of the DMN in the EC state (Figure 10). Furthermore, there was a significant positive correlation between the increase in the mean PVT RT and the enhancement in the effective connectivity from the right precentral gyrus to the left superior frontal gyrus (r = 0.481, p = 0.037) of the SMN in the EO state (Figure 11). However, after FDR correction, neither correlation was significant.

Our study’s fICA and iCoh results showed that in RW, there was a tendency for brain activity changes in the EO state compared to those in the EC state; notably, there was an increase in high-frequency bands, and a decrease in low-frequency bands, for the DMN and SMN. Moreover, the iCoh results showed that in TSD, the effective connectivity in the EO state was stronger than that in the EC state.

In SMN, the neural oscillations in the postcentral gyrus and superior parietal lobule in the EO state were lower in both δ and θ bands and higher in both α and β bands than those in the EC state. Furthermore, the effective connectivity was enhanced in the α, β, and γ bands; this enhancement was most significant in the β band. Prior fMRI studies have demonstrated that activity in the SMN in the EO state is lower than that in the EC state. Liang et al. (2014) observed that compared to the EO state, the EC state induced higher fractional ALFF values in the sensation-related cortex. Furthermore, Liu et al. (2013) demonstrated that a striking similarity among the results of three different methods (ALFF, regional homogeneity, and seed-based correlation analysis) was evident for different eye states. This similarity included the increased visual, decreased primary sensorimotor, and auditory cortices in the EO state than in the EC state. Our study assessed the difference in spontaneous and synchronous activity in the SMN between EO and EC states from a space-frequency perspective, and we found that the decreased activity in the SMN in the EO state was mainly in the low-frequency bands (δ and θ), while the activity in high-frequency bands (α and β) increased. It was reported that the EO state induced lower δ and θ powers of the EEG spectra compared with the EC state (Allen et al., 2018). High-frequency activity is demonstrated to be associated with higher cognitive process (Deligianni et al., 2014; Fries, 2015; Sockeel et al., 2016), and increased alertness is accompanied by higher EEG activation in the β-band (Kamiński et al., 2012). Therefore, the higher activity of the SMN at high frequencies may be related to greater alertness and sensory information input during the EO state compared to those during the EC state.

Neural oscillations of the PCu in the DMN in the EO state were lower in both the δ and θ bands than those in the EC state. Furthermore, effective connectivity in the DMN was higher in the EO state in α, β, and γ bands compared with the EC state. This enhancement was most significant in the β band. Our study’s findings suggest significant differences in both the spontaneous and synchronous activity in the DMN between EC and EO states. This is consistent with previous studies (Yan et al., 2009). The higher activity of the DMN was mainly in the high-frequency (α and β) bands. In the DMN, the PCC is related to the general monitoring of sensory information (Raichle and Snyder, 2007). Furthermore, the PCu, MFG, and lateral parietal cortex have been demonstrated to be involved in episodic memory retrieval (Cavanna and Trimble, 2006). Moreover, the PCC and MFG have been reported to be associated with mind wandering and daydreaming (Mason et al., 2007). Therefore, the higher activity of the DMN in the EO state at high frequency may be due to increased gathering and monitoring of environmental information that caused more mind wandering or daydreaming compared with the EC state (Yan et al., 2009).

Previous studies have found inconsistent patterns of activity change between the EO and EC states in the DMN regions and the visual cortex (Yang et al., 2007; Liang et al., 2014). Our study’s findings indicate that these contradictory change patterns are associated with changes in frequency, with higher activity in EO states at high frequencies and lower activity in EO states at low frequencies.

Furthermore, in TSD, the effective connectivity in the EO state was found to be stronger than that in the EC state both at high and low frequencies. TSD is known to cause cognitive decline and a shortage of cognitive resources (Chee and Chuah, 2007; Simon et al., 2015; Lei et al., 2016). After TSD, microsleep was shown to intrude into wakefulness both in the EC and EO states owing to tremendous sleep pressure even though it was resisted by participants (Goel et al., 2009). However, the duration and frequency of microsleep are increased in the EC state compared to those in the EO state. Additionally, opening eyes could be an intentional behavior to mask the tendency to sleep to some extent, which reflects an antagonistic process exerted by individuals against sleepiness. Therefore, the higher effective connectivity in the EO state compared with that in the EC state in TSD may reflect participants’ efforts to combat sleep stress.

In TSD, the activity of DMN was decreased, while SMN was increased, compared to those of RW. Moreover, the changed effective connectivity in the DMN and SMN after TSD was positively correlated with the increased PVT reaction time.

Several investigations have shown that SD results in lower activity in the DMN than RW and creates higher activity in the SMN. Several fMRI studies observed that SD induced significantly declined FC within the DMN in both resting states and visual attention task (Sämann et al., 2010; De Havas et al., 2012; Lei et al., 2015). Another fMRI study observed the functional connectivity densities (FCDs) at rest were lower in the DMN after TSD, and higher in the SMN. Besides, the declined FCD in the PCC was correlated with the increased RT of PVT after TSD (Yang et al., 2018). Cai et al. (2021) observed significantly lower ALFF of the DMN and frontal–parietal network after SD, and higher ALFF in the visual and motor cortex. Our study’s results regarding the DMN and SMN agree with those observed in previous studies.

The PCu is considered the core node of the DMN. It has been proposed that DMN mediated intrinsic activity after SD largely depends on the PCu and PCC with greater task-related deactivation (Fransson and Marrelec, 2008). Gujar et al. (2010) observed an imbalanced pattern of task-induced deactivation, with significantly less deactivation in the SD group within the dorsal anterior cingulate cortex region of the DMN yet significantly greater deactivation in the PCu. They also demonstrated that the imbalance was more indicative of dysfunction than compensatory effects, and accurately distinguished between sleep-deprived and normal people. Hence, we hypothesized that the decreased neural oscillation in the low-frequency band and decreased effective connectivity in the high-frequency band in the PCu suggested deficits in higher cognitive functions after TSD. Furthermore, that declined activity in the PCu might be a biomarker sensitive to indexing sleep loss.

Decreased visual-parietal FC after SD was observed by Fu et al. (2022) indicating the impaired function of processing environmental stimuli. SD can cause a continuous decrease in vigilant attention (Borragán et al., 2019; Qi et al., 2021). However, the brain compensates for this dysfunction, and this compensatory adaptation can partly offset impaired cognitive function (Drummond et al., 2000). Thus, the increased effective connectivity from the right precentral gyrus to the left superior frontal gyrus may compensate for drowsiness. The positive correlation with decreased PVT performance may indicate that the impaired alertness function was related to the blocked information gathering and disruption of the SMN.

The energy allocation model can also demonstrate our point of view that the limited cognitive resources after SD will be optimally allocated for survival (Schmidt, 2014). The DMN is mostly associated with some higher cognitive functions (Mason et al., 2007). Therefore, it is suppressed after TSD. However, activities of the SMN are vital to survival, with an important role in external information gathering, so the enhancement of SMN activity is considered to be the brain’s attempt to maintain alertness and attention to surroundings (Yang et al., 2018; Cai et al., 2021).

The current study indicated that the effective connectivity in the different networks (EO-EC) of the SMN in the β band was significantly decreased after TSD compared to RW. This indicates that TSD impaired alertness and sensory information input, especially in the EO state. Furthermore, in the different networks of the SMN, the effective connectivity from the left superior parietal lobule to the left precentral gyrus was decreased in the β and γ bands. Furthermore, the effective connectivity from the left postcentral gyrus to the left precentral gyrus was decreased in the γ band after TSD compared to RW. Nevertheless, no significant difference was observed in the neural oscillation of (EO-EC) between the RW and TSD.

The postcentral and precentral gyri are the main areas of the human brain’s motor and somatosensory cortices, respectively. The superior parietal lobule, which makes up a large portion of the parietal lobe, has many functions, such as audio-visual multisensory integration (Hawkins et al., 2013), transformation of sensory inputs into an appropriate motor response (Hecht et al., 2013), and visuospatial attention (Cai et al., 2013).

Additionally, many studies have indicated that β-activity is associated with attention arousal and sensorimotor function (Lalo et al., 2007; Kamiński et al., 2012; Gola et al., 2013). These decreased connections related to sensation and motor function in high-frequency bands jointly indicated that TSD disrupted resting SMN functions more in the EO state than in the EC state. This might cause a greater decline in environmental information input in the EO state. Decreased β activity can also be considered to be related to decreased alertness (Kamiński et al., 2012), and γ oscillations are associated with various cognitive activities, such as high alertness and attentional states (Theriault and Perreault, 2019). Usually, the EO state enhances participants’ level of alertness to their surroundings compared to the EC state. Thus, compared to RW, TSD might suppress information gathering, input process, and the level of alertness in the EO state as opposed to the EC state.

In summary, the SMN’s functions of alertness and collecting the surrounding environment’s information in the EO state were impaired after TSD with excessive drowsiness and limited psychological resources. The results demonstrated that alertness and sensory information input in the EO state were more impaired after TSD than those in the EC state. This implies that the EO state can better reflect the influence of TSD on cognitive function. Thus, it may be better to use an EO state rather than an EC state in future resting-state research.

The present study had some limitations. First, the study detected the effects of different eye states on the DMN and SMN. However, the interactions between different networks remain unclear. Second, given the partial overlap between different functional networks, other networks may have affected the selected network nodes. Third, because our experiment required participants to spend three nights in the lab, and we only had male nursing staff for continuous monitoring of participants, so, we did not recruit women, which might have induced bias in the results. It would be interesting to study gender differences in the effects of TSD on resting brain activity in the future. Fourth, the sample size was small. Fifth, the effective connectivity we measured is a form of static functional connectivity, but the internal mental activity of participants varies with time. Hence, future research should observe dynamic functional connectivity to identify the difference between EC and EO states. Sixth, the correlations between changes in effective connectivity and changes in PVT reaction time were not significant after multiple comparisons correction, which may be due to the small sample size. Additionally, the RW data were acquired at 8:00, while the TSD data were acquired at 20:00 on the next day. This might have mixed the effects of the circadian rhythm. Plus, the small sample size may also be a contributing factor. Thus, future studies should further investigate the differences in the interactions of networks between different eye states. In addition, the different effects of TSD on functional networks in different eye states should be further studied.